If we suppose we have an entangled pair in position/momentum and, following the argument, we measure position of particle A. We get a result let say xA. Then we want to predict the measurement of position of B, so up to now we have not measured particle B, but we know it's wave-function is a delta in -xA. Since we have to carry the information from A to B, the wave-function in B evolved into a gaussian spreading wave-packet. Then we measure the position of B and we get xB which is now not forcedly -xA even if it is the most probable. So that in this case, we cannot predict the outcome of measurement of B with certainty (or exactness) when we compare with measurement in A. If we program the measurement time of A and B, they have to be simultaneous, but this would mean relatively to a reference frame. If let say one frame of measurement is moving, then the events are no more simultaneous depending on the frame, so that the wave-function spread out again according to Schroedinger's equation, hence the prediction with certainty is not possible neither ? Even if for the latter case I don't really understand in which frame the evolution arises. Maybe somebody could clear it up, thanks.